Projects per year
Abstract
Quality assurance remains a significant challenge for laser powder bed fusion and metal additive manufacturing. Despite system manufacturers offering process monitoring as a possible solution, datasets are large and cumbersome with practical use limited without direct comparative data. Model datasets would enable individual build validation, highlight deviations, and facilitate intelligent build planning whereby challenging features or build strategies could be pre-emptively assessed.
Herein a pragmatic approach has been developed to model process monitoring data from a commercial system using a relatively simple algorithm. Using a heuristic method, the algorithm response has been fitted to an experimental dataset to derive governing constants and their relationship to key process parameters. Predictability of constants and model fit has been shown to improve with increasing line energy up to a maximum R2=0.8. Algorithm variable trends, supported by corresponding sensitivity analysis, identified two different behavioural regimes. Under low linear energy density (<0.2J/mm) the cumulative spacetime proximity time-weight variable shows a low sensitivity index, characterised by a flat model response reflected in the experimental data. At higher energies (≥0.2J/mm) algorithm variables become more predictable, reflected in stabilising sensitivity indices, as measurements adopt a form characteristic of the cumulative spacetime proximity function.
Effectiveness has been demonstrated through presentation of experimental and model data. Refining the methodology to accommodate noise, geometry, and systematic behaviours are identified as key steps to future development. This feasibility study has laid the groundwork for a generalised predictive tool, capable of realising the quality assurance ambitions promised by laser powder-bed fusion process monitoring.
Herein a pragmatic approach has been developed to model process monitoring data from a commercial system using a relatively simple algorithm. Using a heuristic method, the algorithm response has been fitted to an experimental dataset to derive governing constants and their relationship to key process parameters. Predictability of constants and model fit has been shown to improve with increasing line energy up to a maximum R2=0.8. Algorithm variable trends, supported by corresponding sensitivity analysis, identified two different behavioural regimes. Under low linear energy density (<0.2J/mm) the cumulative spacetime proximity time-weight variable shows a low sensitivity index, characterised by a flat model response reflected in the experimental data. At higher energies (≥0.2J/mm) algorithm variables become more predictable, reflected in stabilising sensitivity indices, as measurements adopt a form characteristic of the cumulative spacetime proximity function.
Effectiveness has been demonstrated through presentation of experimental and model data. Refining the methodology to accommodate noise, geometry, and systematic behaviours are identified as key steps to future development. This feasibility study has laid the groundwork for a generalised predictive tool, capable of realising the quality assurance ambitions promised by laser powder-bed fusion process monitoring.
Original language | English |
---|---|
Article number | 100252 |
Number of pages | 13 |
Journal | Additive Manufacturing Letters |
Volume | 11 |
Early online date | 29 Oct 2024 |
DOIs | |
Publication status | Published - Dec 2024 |
Keywords
- Additive manufacturing
- Process monitoring
- Laser powder bed fusion (LPBF)
- process modelling
Fingerprint
Dive into the research topics of 'Modelling process monitoring data in laser powder bed fusion: A pragmatic route to additive manufacturing quality assurance'. Together they form a unique fingerprint.-
Rapid Design of Bioinspired Alloys - From Modelling to Manufacture
Mottura, A. (Co-Investigator), Cox, S. (Principal Investigator) & Knowles, S. (Co-Investigator)
15/11/21 → 31/10/25
Project: Research Councils
-
Invisible Customisation - A Data Driven Approach to Predictive Additive Manufacture Enabling Functional Implant Personalisation
Cox, S. (Principal Investigator) & Andrews, J. (Co-Investigator)
Engineering & Physical Science Research Council
1/05/20 → 30/04/23
Project: Research Councils